Nonsteroidal antiinflammatory drugs (NSAIDs) are currently being assessed for use in the prevention and treatment of colorectal cancer. Their promise is based on epidemiologic, preclinical and clinical evidence backed up by in vitro mechanistic studies.1, 2
A property of all NSAIDs is their ability to inhibit cyclooxygenase (COX). COX catalyses the first step in the formation of prostaglandins from arachidonic acid. Two isoforms of COX are known to exist: COX-1 and COX-2. In contrast to COX-1, COX-2 is not expressed constitutively in most tissues, including the colon and rectum, but is induced by a variety of inflammatory mediators, cytokines and growth factors.3 NSAIDs may be classified in terms of their selectivity in inhibiting one COX isoform over another. Classical NSAIDs, such as sulindac and aspirin, are equipotent or selective for COX-1, whereas new NSAIDs such as celecoxib that selectively inhibit COX-2 have been developed. These have reduced gastrointestinal toxicity.4
Consistent with the antineoplastic effect of NSAIDs, COX-2 is expressed constitutively in the tumour epithelium of colorectal carcinomas and experiments using human colorectal carcinoma-derived cell lines show that epithelial COX-2 promotes angiogenesis, tumour cell invasion and resistance to apoptosis.5, 6, 7, 8 These indications that COX-2 has tumorigenic function reinforce studies using murine models of intestinal neoplasia; most notably, a COX-2 null mutation dramatically reduced the number and size of intestinal tumours that develop in ApcΔ716 mice.9
There is also evidence that NSAIDs have anticancer effects independent of the inhibition of COX-2 and resultant reduction in prostaglandin (PG) production. These include the inhibition of cell cycle progression and the induction of apoptosis (reviewed in refs. 2 and 10). The induction of apoptosis, in particular, is an important antitumour effect of NSAIDs. In vivo, NSAIDs induce apoptosis of intestinal tumour cells, which correlates with their chemopreventive efficacy.11, 12, 13 In cultured colorectal tumour cells, the induction of apoptosis is the dominant antiproliferative effect of sulindac sulphide and salicylate (which inhibit both COX isoforms) and of COX-2-selective NSAIDs such as celecoxib and NS-398 (reviewed in ref. 10). These studies have utilised concentrations of NSAID greater than that required to inhibit prostaglandin synthesis fully; hence it is not clear to what extent the induction of apoptosis is dependent on the inhibition of the COX isoforms. It occurs in colorectal carcinoma cells lacking COX-1/-2 expression,14, 15 but there is also evidence that COX inhibition may be involved.16, 17 The use of these higher concentrations of NSAID is justifiable, as the efficacy of NSAIDs in cancer therapy is actively being evaluated and in this context there is scope for achieving higher serum concentrations of NSAIDs. In addition, this approach may highlight key pathways regulating the proliferation and apoptosis of carcinoma cells.
Elucidation of the pathways that signal the effects of NSAIDs on colorectal tumour cells is at an early stage. Various NSAIDs, including the highly selective COX-2 inhibitor NS-398, upregulate COX-2 protein expression in colorectal carcinoma cell lines whilst inhibiting its activity.18, 19, 20 We have found that NS-398 simultaneously inducs apoptosis in such cell lines.18 As the induced expression of COX-2 protein in response to various stimuli in several cell types, other than colonic epithelia, is known to be regulated by mitogen-activated protein kinase (MAPK) cascades,21, 22, 23 we hypothesised that one or more of the MAPK cascades may mediate the effects of NS-398 on colorectal tumour cells. In the present report, we show that NS-398 causes sustained activation of ERKs that results in the upregulation of COX-2 protein and the induction of apoptosis in colorectal carcinoma cells.
MATERIAL AND METHODS
Cell culture and treatment
The HT29 colon carcinoma cell line was used in this study. Cells were grown as an adherent monolayer in T25 tissue culture flasks in DMEM supplemented with 10% FBS (but see below), penicillin (100 U/ml), streptomycin (100 μg/ml) and glutamine (2 mM).18 The highly selective COX-2 inhibitor NS-398 (Cayman, Ann Arbor, MI) was made up as a 30 mM stock solution in DMSO and stored at −20°C. The inhibitor of MEK-1/-2 activation, UO126 (Tocris Cookson, Ballwin, MO) was made up as a 10 mM stock solution in DMSO, stored at −70°C and used within 1 month. U0126 has little to no effect on kinases other than MEK-1/-2. (24). Anisomycin (CN Biosciences, Nottingham, UK) was prepared as a 50 mg/ml stock solution in DMSO and stored at −20°C. For treatment with NS-398 and/or U0126, cells were harvested by trypsinisation (0.05% [w/v] trypsin/EDTA), seeded at 106 cells per flask and cultured for 3 days. Cell cultures were treated in triplicate with 75 μM NS-398 and/or 1 μM to 2.5 μM U0126, for up to 96 hr. Each control and treated flask received an equal final volume of DMSO, to a maximum of 0.28% (v/v). At these concentrations, DMSO alone did not affect cell viability (data not shown). NS-398 was used at 75 μM as this is the approximate inhibitory concentration at 50% (IC50) value for this compound when HT29 cultures are treated (measured by attached cell yield at 96 hr) and this concentration induces apoptosis.18 This dose of NS-398 is greater than that required to inhibit PGE2 synthesis fully in the HT29 cells,18 thus allowing the effects of NS-398 to be independent of the inhibition of COX-2. Routinely, when both agents were used, cells were preincubated with the required concentration of U0126 for 2 hr before combined treatment with U0126 and NS-398. However, equivalent results were obtained with or without this pretreatment. Also, results equivalent to those described were obtained when NS398 (75 μM) and U0126 (1 and 2.5 μM) were applied to the cells in medium that was not supplemented with FBS.
Treatment with anisomycin was at 300 μg/ml for 1 hr. At the indicated times, cell lysates were prepared (see below) or the attached cells (those that adhered to the tissue culture flask) and the floating cells (those that detached from the adhered monolayer) were counted separately using a haemocytometer. Cell counts were performed at 96 hr to reflect effects of molecular events occurring until 72 hr. Separate aliquots of attached and floating cells and aliquots of the pooled attached and floating cell populations were stained with acridine orange and ethidium bromide for analysis by fluorescence microscopy (see below).
Measurement of apoptosis
The extent of apoptosis in the cultured epithelial cell lines was assessed for each experiment by (i) measuring the fraction of the adherent cells that were morphologically apoptotic; (ii) measuring the proportion of cells that had detached from the cell monolayer and were floating in the medium and determining the fraction of these floating cells that were morphologically apoptotic; and (iii) measuring the proportion of cells with apoptotic morphology in aliquots of combined attached and floating cell populations. Morphologic assessment of apoptosis after staining with acridine orange and ethidium bromide was as described.18 For NS-398 and/or U0126 treatment, the proportion of adherent cells showing apoptotic morphology did not increase above control (spontaneous) levels, whereas the percentage of floating cells showing apoptotic morphology in treated cultures was high (>85%) and similar to that in control cultures. This meant that in a given culture the percentage of the total cell population that was floating reflected the extent of apoptosis. Data obtained in this way were in close agreement with data obtained directly from morphologic analysis of attached and floating cell populations when combined. The nature of cell death was also investigated by examining the cleavage of poly (ADP-ribose) polymerase (PARP) in floating cells from control and treated cultures. PARP was detected by SDS-PAGE immunoblotting (see below).
Cell extracts and SDS-PAGE immunoblotting
Cell cultures were washed twice with ice-cold PBS and then lysed in situ for 20 min at 4°C in buffer consisting of 20 mM Tris-HCl (pH7.5), 150 mM NaCl, 1 mM EDTA, 1% Triton-X-100, 2.5 mM sodium pyrophosphate, 1 mM β-glycerophosphate and 1 mM sodium orthovanadate, supplemented with a protease inhibitor cocktail for mammalian cell extracts (Sigma, Poole, UK). The cell lysate was then vortexed vigorously for a total of 60 sec, with intermittent cooling on ice and centrifuged for 10 min at 10,000g and 4°C. Protein assays (Detergent Compatible Protein Assay, Bio-Rad, Hemel Hempstead, UK) were carried out on aliquots of this supernatant, which were stored at −70°C prior to use. Prior to loading, aliquots of cell lysate supernatant (100–150 μg protein) were mixed with one-fifth their volume of 5× loading buffer (0.3 M Tris [pH 6.8], 10% [w/v] SDS, 50% [v/v] glycerol, 25% [v/v] 2-mercaptoethanol). Where indicated, rather than lysis in situ, cells were trypsinised and lysates of 2 × 106 cells prepared as described previously.18 Western blotting was also as described.18
COX-2 protein was detected using affinity-purified polyclonal anti-cyclooxygenase-2 at 1 μg/ml (a gift from Dr. Stephen Prescott, University of Utah, Salt Lake City, UT). PARP was detected using a monoclonal antibody (at 1:2,000 dilution) that recognises full-length PARP (116 kDa) and the 85 kDa apoptosis-induced cleavage product (Alexis Biochemicals, Nottingham, UK).
MAPK proteins and their phosphorylated (active) forms were detected with antibodies from New England Biolabs (Beverly, MA), used at the dilutions recommended by the manufacturer: ERK1 and ERK2 (#9102), phosphorylated ERK1 and ERK2 (#9101L), p38 (detects α-p38, #9212), phosphorylated p38 (detects phospho-α, β, γ, δ-p38, #9211S), p54/p46 JNK (#9252), phosphorylated JNK (#9251S), p90RSK (#9342), phosphorylated p90RSK (#9341). When cell lysates were prepared from pellets of trypsinised cells, loading and transfer were controlled by repeat probing with anti-α-tubulin (Sigma) at 1:10,000 dilution. Membranes were subsequently probed with anti-rabbit antibody or anti-mouse antibody (for α-tubulin) conjugated to horseradish peroxidase (1:1000; Sigma). Bound antibodies were detected using enhanced chemiluminescence (Kirkegaard and Perry, Gaithersburg, MD). If necessary, membranes were stripped by incubation at 50°C for 30 min in a solution of 62 mM Tris-HCl, pH 6.7, 2% SDS and 100 mM 2-mercaptoethanol.
Statistical analysis was carried out using SPSS for Windows statistical software (release 10.0.5, SPSS Inc. Chicago II). Two-way analysis of variance (ANOVA) was used to determine differences amongst means. Pairwise comparisons were made using Tukey's post hoc test for multiple comparisons.
The COX-2-selective inhibitor NS-398 causes sustained activation of ERKs in HT29 cells
Previously, we have shown that NS-398 upregulates COX-2 protein expression, whilst abolishing its activity, concurrent with inducing apoptosis and growth inhibition of colorectal tumour cells.18 To understand further the mechanisms underlying these effects, we analysed whether NS-398 activated the MAPKs ERK, p38 and JNK. HT29 cells were treated with 75 μM NS-398 for up to 72 hr and the activity of ERK, p38 and JNK was assessed using monoclonal antibodies that recognise only the catalytically active, phosphorylated forms of the kinases. NS-398 caused activation of ERK1 and ERK2 (p44 and p42, respectively) that was reproducibly seen from 24 to 72 hr (Fig. 1a) but not between 30 min and 10 hr (data not shown). Total ERK-1/-2 (phosphorylated and nonphosphorylated) levels were not affected by treatment with NS-398 and were therefore used to control for ERK activation. In contrast to ERK-1/-2, the phosphorylation status of p38, determined between 30 min and 72 hr, was not affected by treatment with NS-398 (Fig. 1a and data not shown). No phospho-JNK was detectable in control or treated cultures. (Fig. 1a illustrates total JNK expression.) Figure 1b indicates that the phosphorylation of p38 and JNK was detectable in HT29 cells when they were treated with anisomycin.
The MEK1/2 inhibitor U0126 prevents NS-398-induced expression of COX-2 protein
The activation of ERK-1/-2 and the upregulation of COX-2 protein expression are seen at 24 hr, but not at 10 hr, after treatment with NS-39818 (Figs. 1, 2). This timing allowed the ERK pathway to be involved in the upregulation of COX-2 protein induced by NS-398. To determine whether this was the case, ERK activation was inhibited using U0126, a selective inhibitor of the ERK-activating kinases MEK-1/-2. At 1 and 2.5 μM, U0126 inhibited the activation of ERK-1/-2 induced by 75 μM NS-398 (Fig. 2). The inhibition was dose-dependent and maintained active ERK-1/-2 approximately at the control level. The protein kinase p90RSK, a substrate of ERK-1/-2, was also found to be phosphorylated by treatment with NS-398. Its phosphorylation, like that of ERK-1/-2, was inhibited by U0126. This finding provides further evidence of the increased activity of ERK upon NS-398-induced phosphorylation and indicates that U0126 is acting by inhibition of the MEK-ERK pathway. Concomitant with the inhibition of ERK activity, U0126 prevented the upregulation of COX-2 protein expression that was induced by NS-398. Again, the effect was dose-dependent and COX-2 was maintained at the control level by 2.5 μM U0126. U0126 did not affect the expression of total ERK-1/-2 or of total p90RSK (Fig. 2). Treatment with U0126 did not alter the phosphorylation status of p38 or JNK (data not shown).
Prevention of ERK activation protects HT29 cells from the antiproliferative effect of NS-398
Since the ERK pathway may affect cell proliferation and apoptosis,25, 26 we investigated whether the activation of ERK-1/-2 by NS-398 regulated the antiproliferative effect of this NSAID. We have shown previously that the effect of NS-398 on cell yield and apoptosis are first detectable at approximately 48 hr and hence follow the activation of ERK observed in the present study18 (Fig. 1). HT29 cells were treated for 96 hr with 75 μM NS-398, alone and combined with U0126 to inhibit NS-398-induced signalling through ERK-1/-2. The results are illustrated in Figure 3. Consistent with previous findings,15, 18 NS-398 caused a reduction in attached cell yield and increased the extent of cell death that was morphologically characteristic of apoptosis (not shown). Floating cells from control and treated cultures contained only the 85 kDa form of PARP that is characteristic of apoptotic cells (Fig. 3b). U0126 protected cells from the effect of NS-398, inhibiting, in a dose-dependent manner, the reduction in attached cell yield and the induction of cell death (statistically significant; see Fig. 3a). The effect of U0126 occurred at doses that inhibited NS-induced ERK activation (Fig. 2) but alone had no significant affect on cell yield or cell death (Fig. 3a).
The results presented in our study indicate a key role for the MEK/ERK pathway in apoptosis and in the regulation of COX-2 protein expression induced by a COX-2-selective NSAID (NS-398). This is based on the following observations: treatment with NS-398 results in sustained MEK-dependent activation of ERK that precedes a decrease in cell yield and induction of apoptosis and also precedes or accompanies an increase in COX-2 protein expression. These effects of NS-398 are inhibited when NS-398-induced ERK activation is prevented by inhibition of MEK.
One interesting aspect of these results is the pro-apoptotic role played by ERK in NS-398-induced apoptosis. The activation of ERK is classically associated with cell survival and proliferation.25, 26 However, recent studies have indicated that this is not always the case and that apoptosis or cell cycle arrest may result from ERK activation.27, 28, 29, 30, 31 The cell fate in response to signalling through ERK depends on the nature, strength and duration of the stimulus and on the cell type. Our data indicate that when colorectal carcinoma cells are treated with NS-398 there is a sustained activation of ERK and this signals apoptosis. The results with NS-398 are consistent with other studies indicating a pro-apoptotic or growth-inhibitory role for ERK.27, 29, 31 For example, cannabinoids induce apoptosis of C6 glioma cells, which is dependent on ERK activation that is detectable after 4 days of treatment.29 It is notable that although ERK is strongly activated in response to NS-398, there was no detectable phosphorylation of the p38 or JNK kinases that are more commonly associated with the response to an apoptotic stimulus or to nonspecific cellular stresses.
The data presented here indicate that induced ERK activity, as well as mediating NS-398-induced apoptosis, is required for the NS-398-induced increase in COX-2 protein expression, thus indicating that these effects of NS-398 are not independent of each other. Indeed, it is likely that the induced expression of COX-2 is a response to an apoptotic stimulus.8, 22 There is evidence that in HT29 cells treated with sodium butyrate this response constitutes a form of induced chemoresistance,8 thus highlighting the potential of COX-2 selective inhibitors as therapeutic adjuvants. In the case of induced expression of COX-2 by NSAIDs, the activity of the cyclooxygenase component of COX-2 is inhibited concomitant with its upregulation.
Of particular note with respect to the present study is that ceramide treatment of human mammary epithelial cells induces COX-2 protein expression, which is prevented by inhibition of MEK.22 This may imply a role for ceramide accumulation in NS-398-induced apoptosis, as has been suggested for other NSAIDs17 and is consistent with the delayed activation of ERK that we observe. Whether the induction of COX-2 protein by NSAIDs reduces their efficacy as anticancer agents is not known. It may lead to a transient increase in COX-2 activity when NSAID use is discontinued and would increase the expression of the potentially mutagenic peroxidase component of this bifunctional enzyme.18, 19 However, NSAIDs inhibit cytokine-induced COX-2 expression,20 which may be a more significant effect in vivo.
Downstream effectors of ERK that potentially mediate its COX-2 regulatory and pro-apoptotic effect after activation by NS-398 include the NF-κB signalling pathway,32 which may be pro-apoptotic when activated by the MEK/ERK pathway28 and may also induce COX-2 expression;33;33 and the pro-apoptotic protein Par-4, which is induced by high concentrations of NS-39834 and is regulated in expression by a MEK-dependent pathway.35
Our study establishes that the MEK/ERK signalling cascade mediates the induction of apoptosis and of COX-2 protein expression in colorectal carcinoma cells treated with a COX-2-selective NSAID. Studies such as ours (and the others discussed) specify intracellular signalling pathways that underpin the cellular response to NSAIDs. Such knowledge will allow a more selective and disease-specific approach to combat the genesis and recalcitrance of colorectal cancer.
We are grateful to Drs. C. Stewart and A. Hague for helpful discussions and to Dr. S. Prescott for providing the COX-2 antibody.